Master thesis in Environmental & Infrastructure Planning
Finding Routes for Hyperloop Transportation Infrastructure in the Netherlands Using a GIS-
MCDA Approach
by
Niek Bebelaar
n.bebelaar@student.rug.nl
s2206056
Faculty of Spatial Sciences University of Groningen
Supervisor: Dr. T. Busscher Second reader: Drs. A.M. Hilbers
January 2020
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
_____________________________________________________________________________________________________________________________________________________________________
A B S T R A C T
Plans exist for a Hyperloop (test) track in the Netherlands. Hyperloop is an innova- tive transportation technology with a dedicated tube-infrastructure.
Currently there is no suitable institutional framework for the implementation of Hyperloop in the Netherlands. Therefore, the GIS-MCDA approach is chosen be- cause this location problem can be defined as a multi-criteria decision problem, whereby a variety of spatial criteria have to be taken into account and various stake- holders are involved. The research question is defined as follows: where to place the route of Hyperloop infrastructure in the Netherlands taken into account an origin and destination location, using a GIS-MCDA approach?
The area-oriented planning approach is used, whereby interaction of infrastruc- ture with its surrounding is considered. From the current configuration of the study area, expressed in 17 criteria, are routes calculated. With a Pairwise Comparison questionnaire (n=16) consulting experts were importance values assigned to criteria.
Five response groups were defined, each with different sets of weights.
The result is 10 possible routes of which most routes covered the same area, had similar scores for evaluation statistics (accumulated cost, length, sinuosity) and they crossed four identified bottlenecks at the same location. Integration of the method with the Dutch Trac´ewet infrastructure planning procedure is also discussed.
Suggestions for future work include an analysis of the effect of buffer sizes around specific features, the use of a (non-spatial) MCA for choosing a “best” route, and the automation of parts of the methodology in order to make it an iterative process for a workshop setting.
Keywords: Hyperloop, GIS-MCDA, least cost path analysis, infrastructure route design, transportation infrastructure, area-oriented planning
iii
C O N T E N T S
1 introduction 3
1.1 Motivation . . . 3
1.2 Infrastructure developments . . . 4
1.3 Multi criteria decision problems . . . 5
1.4 Research objective and research questions . . . 6
1.4.1 Research objective . . . 6
1.4.2 Research questions . . . 7
1.5 Reading guide . . . 8
2 theoretical framework 9 2.1 The Hyperloop . . . 9
2.2 Institutional design . . . 10
2.3 Frameworks for decision making . . . 11
2.4 Core concepts of the GIS-MCDA approach . . . 13
2.4.1 Spatial simulation and spatial optimization . . . 13
2.4.2 Value scaling, criteria weighting, and the combination rule . . 13
2.5 Literature reviews on GIS-MCDA . . . 14
2.6 Examples of GIS-MCDA in infrastructure route planning . . . 15
2.6.1 Applications with the simulation approach . . . 15
2.6.2 Applications with the optimization approach . . . 18
2.6.3 Involvement of stakeholder groups . . . 19
2.7 Least cost path analysis and its use in various applications . . . 19
2.8 Conclusion theoretical framework . . . 21
3 selection of criteria 25 3.1 Requirements for criteria . . . 25
3.2 Selected criteria . . . 26
3.3 Preprocessing of datasets . . . 30
3.4 Conclusion selection of criteria . . . 30
4 methodology 33 4.1 Proof of Concept for methodology . . . 33
4.2 Study area for Hyperloop in The Netherlands . . . 33
4.3 Criteria from the environment regarding Hyperloop infrastructure . . 34
4.4 Requirements of Hyperloop infrastructure . . . 35
4.4.1 Curvature of the track . . . 36
4.4.2 Above or below ground . . . 36
4.4.3 Acceleration of the vehicles . . . 37
4.4.4 Proof of Concept: requirements of Hyperloop infrastructure . 37 4.5 Importance of criteria with Pairwise Comparison . . . 37
4.6 Rasterization . . . 42
4.6.1 Cell size . . . 43
4.6.2 Buffer size . . . 43
4.7 Cost Surface maps . . . 45
4.8 Calculating a route from a Cost Surface . . . 45
4.9 Evaluation of calculated Least Cost Paths . . . 46
4.10 Least Cost Path that obeys requirements of Hyperloop infrastructure . 47 4.11 Conclusion methodology . . . 48
5 results and discussion 49 5.1 Questionnaire invitations and responses . . . 49
v
5.2 Cost Surfaces on the map . . . 49 5.3 Least Cost Paths . . . 53 5.4 Integration with Dutch practice of designing transportation corridors 58 5.5 Drawbacks of this methodology . . . 59
6 conclusion 63
6.1 Answer main research question . . . 63 6.2 Reflection on the research . . . 64 6.3 Recommendations for future work . . . 64
a selection procedure for criteria 71
b questionnaire for weighting of criteria 73
c bag gebruiksdoel category per defined criterion 79
d processing models in gis software 83
L I S T O F F I G U R E S
Figure 1.1 Impression of a Hyperloop station. Image created by Hardt Hyperloop. . . 3 Figure 1.2 Corridor for the OV Schiphol-Amsterdam-Almere-Lelystad (SAAL)
project. The project contains improvements and upgrades of the existing rail roads in the area (visualized in green). . . 4 Figure 2.1 Future cross section of the variety of transportation infras-
tructure technologies? Design for the Hyperloop infrastruc- ture is visualized in pink. Image fromVan Lint[2019]. . . 11 Figure 2.2 Impedence maps for four visions: equal, social, ecology, and
economy [Keshkamat et al.,2009]. These type of raster maps are input for a least-cost path search procedure. . . 17 Figure 2.3 Traditional approach to Least Cost Path Analysis. Image
fromStefanakis and Kavouras[1995]. . . 20 Figure 2.4 Conceptual model . . . 23 Figure 3.1 Line-oriented versus area-oriented criteria. Figure created by
the author and inspired by Heeres et al. [2012]. The area- oriented approaches in (road) infrastructure planning com- bine two perspectives: inside-out and outside-in, which is also shown in this figure. In this research, the shortlist of criteria should be involved with matters regarding the area surrounding the Hyperloop infrastructure, i.e. area-oriented criteria. . . 26 Figure 3.2 Example of the ArenA: pand layer and verblijfsobject layer. . 27 Figure 3.3 All 17 individual layers fromTable 3.1in the study area. . . . 31 Figure 4.1 The study area with the airports of Lelystad and Schiphol.
The shortest possible path between those locations is 54 kilo- meters. Included on the map is the search area for the test track fromArup et al.[2017]. . . 34 Figure 4.2 Study area for the Proof of Concept, which is used to test
(parts of) the methodology. . . 35 Figure 4.3 A train travels with speed V through the curve, i.e. over the
arch of an (imaginary) circle, with center O and curve bend radius R. Image from [Huston,2017]. . . 36 Figure 4.4 The minimum curve bend radii for the Hyperloop illustrated
in the study area. For vehicle speeds up to 480 km/h the radius is 1.6 km, and for speeds up to 1220 km/h the radius is 4.8 km [SpaceX,2013]. . . 37 Figure 4.5 Hyperloop test track in the Nevada desert, USA. Image from
Virgin Hyperloop One. . . 38 Figure 4.6 The fundamental scale of absolute numbers for pairwise com-
parison. Figure fromSaaty[2008]. . . 39 Figure 4.7 Example of how the ”suggested modification” feature from
the Excel template of Goepel [2013] leads to consistent re- sponses. . . 41 Figure 4.8 Screenshot of the Google Forms questionnaire. . . 42
vii
Figure 4.9 Map 1: Offices A, B and C as vector features.
Map 2: Rasterized map, without a buffer applied. Offices A and B are still included on the map, as black pixels, while office C disappeared.
Map 3: Rasterized map, with a buffer applied. Offices A, B and C are all still visible on the map.
Map 4: Shows all features from this example: the vector fea- tures in purple, the 25m buffer around it, and the rasterized
layer. . . 44
Figure 4.10 Creation of a Cost Surface map from individual raster layers. Afterwards, the Least Cost Path is calculated from the Cost Surface. . . 45
Figure 4.11 The variety of possible Least Cost Paths for the Hyperloop, taking the parameters into account as defined in this research (various weights for criteria, the pixel size of the rasters, and buffer size around ”building” type criteria). The dots repre- sent an outcome: LCP Route and LCP Cost. The colors of the dots clarify which weight group is applied to the resulting LCP. 47 Figure 5.1 These bar charts indicate the consolidated weights per crite- rion and per response group (all equal weights, municipali- ties, provinces, national government, research). . . 50
Figure 5.2 Cost surfaces when various weights are applied: All equal weights, weights from municipalities, and weights from provinces. 51 Figure 5.3 Cost surfaces when various weights are applied: weights from national governmental institutions and weights from research institutions. . . 52
Figure 5.4 Least Cost Paths for five weight groups. . . 54
Figure 5.5 Least Cost Paths for five weight groups, including detail maps at locations discussed inSection 5.2. . . 55
Figure 5.6 Least Cost Paths and population density, south of Amsterdam. 57 Figure 5.7 Effect of buffers around building features on the resulting Least Cost Paths. . . 57
Figure 5.8 Computing time for calculating the Least Cost Paths using a high-performance computer. If cell size (horizontal axis) de- creases computing time (vertical axis) increases exponentially. 60 Figure B.1 Questionnaire page 1 of 4 . . . 74
Figure B.2 Questionnaire page 2 of 4 . . . 75
Figure B.3 Questionnaire page 3 of 4 . . . 76
Figure B.4 Questionnaire page 4 of 4 . . . 77
Figure D.1 The semi-automated process for transforming vector files into raster files. . . 84
Figure D.2 Workbench for calculating weighted raster layers (FME) . . . 85
Figure D.3 Workbench for calculating Cost Surface maps from individ- ual layers (FME) . . . 86
L I S T O F T A B L E S
Table 2.1 Overview of advantages and disadvantages for the Nether- lands and developing countries of being a ”first-mover” with regard to the Hyperloop. Source:Arup et al.[2017]. . . 10 Table 2.2 MCA and CBA compared. Source: Beria et al.[2012]. . . 12 Table 2.3 Overview of stakeholder groups identified in the reviewed
articles on Geographical Information System(s) (GIS)-Multi- Criteria Decision Analysis (MCDA) for (transportation) infras- tructure planning. . . 19 Table 2.4 Overview of factors that are used in the reviewed articles on
GIS-MCDAfor (transportation) infrastructure planning. . . 22 Table 3.1 Overview of selected criteria. . . 30 Table 4.1 Example of the Pairwise Comparison questionnaire for the
Proof of Concept. . . 39 Table 4.2 Random consistency index RI for various amounts of factors
n included in the pairwise comparison. Source:Saaty[1987]. 40 Table 4.3 Raster cell size and processing times. . . 43 Table 5.1 Least Cost Path results for various parameter values. . . 53 Table A.1 Procedure for selecting criteria . . . 72 Table C.1 Overview of BAG gebruiksdoel categories per defined crite-
rion (1 of 3) . . . 80 Table C.2 Overview of BAG gebruiksdoel categories per defined crite-
rion (cont’d, 2 of 3 . . . 81 Table C.3 Overview of BAG gebruiksdoel categories per defined crite-
rion (cont’d, 3 of 3) . . . 82
ix
A C R O N Y M S
AHP Analytic Hierarchy Process . . . .14
CBA Cost Benefit Analysis . . . .11
DEM Digital Elevation Model . . . .43
GIS Geographical Information System(s) . . . .ix
LBCS Line-Based Cartographic Simplification . . . .21
LCP Least Cost Path . . . .34
LCPA Least Cost Path Analysis . . . .15
MAMCA Multi-Actor Multi-Criteria Analysis . . . .14
MCA Multicriteria Analysis . . . .11
MCDA Multi-Criteria Decision Analysis . . . .ix
MCDM Multi-Criteria Decision Making . . . .14
MIRT Meerjarenprogramma Infrastructuur, Ruimte en Transport . . . .5
PCA Principal Component Analysis . . . .18
PCM Pairwise Comparison Matrix . . . .40
SDSS Spatial Decision Support System . . . .6
SMCA Spatial Multi Criteria Analysis . . . .16
SVIR Structuurvisie Infrastructuur en Ruimte . . . .4
1
1 I N T R O D U C T I O N
1.1 motivation
In contemporary transportation planning, sustainable mobility is an ’alternative paradigm within which to investigate the complexity of cities, and to strengthen the link between land use and transport’ [Banister,2008]. Encouraging modal shift (i.e. from the use of cars for transportation to sustainable modes for transportation) and greater efficiency of the transportation system are both actions that contribute to sustainable mobility [Banister,2008]. Maglev trains, biofuel powered vehicles, battery electric vehicles, hydrogen fuel cell vehicles, and e-bikes are exemplar sus- tainable forms of mobility [Leibowicz,2018].
The Hyperloop is a proposed technology for the transportation of people and goods [Arup et al.,2017]. It is proposed by the founder of SpaceX, Elon Musk, and introduced as a fifth mode of transport, next to cars, trains, planes, and boats. The Hyperloop technology uses fast (in theory up to 1220 km/hour) vehicles floating on a magnetic cushion in tubes with low air pressure, so that resistance to movement of the vehicles is reduced [Dudnikov,2017;SpaceX,2013].
With regard to sustainable mobility, Hyperloop could be an interesting devel- opment. Hyperloop could encourage passengers to shift modality if it is well in- tegrated with other forms of public transportation, such as train stations or bus stations. Moreover, Hyperloop could encourage greater efficiency in the transport system if vehicles can leave the stations as regularly as planned (every 2 minutes) [SpaceX,2013].
Figure 1.1: Impression of a Hyperloop station. Image created by Hardt Hyperloop.
The Dutch Ministry of Infrastructure and Water stated the ambition to ”pioneer public transport innovation”, as mentioned in the report ”Public Transport in 2014 - Outlines of a vision for the future” [Ministry of Infrastructure and Water,2019].
Besides, as ”one of the leading countries in the development and implementation of new and innovative mobility concepts” [Arup et al., 2017], the ambition of the
3
Dutch Ministry of Infrastructure and Water is to ”further strengthen and broaden this position” [Arup et al.,2017].
In that regard, the Dutch government conducted a study to the suitability of the construction of a test track for the Hyperloop in the Netherlands [Arup et al., 2017]. The study byArup et al.[2017] resulted in an advice to create a test track of three kilometers in the province of Flevoland. More recently, the Dutch company Hardt Hyperloop announced their plans for building the “European Hyperloop Center” near the city of Groningen, which includes a three kilometer test track [Hardt Hyperloop,2019].
The main reason forArup et al.[2017] for choosing the location in Flevoland was that it might become part of a future commercial route between Schiphol Airport and Lelystad Airport. If the tests with the three kilometers track succeeds, the Hyperloop will be tested on a 40 kilometers track before it can be used to transport people and goods. It is, however, unclear where in the Netherlands this 40 kilometer track should be located. Where in the Netherlands, between Schiphol Airport and Lelystad Airport, should the track (or route or corridor) of this infrastructure be located?
1.2 infrastructure developments
In general, the planning of any (public) transportation project starts with the recog- nition of a current or future need to meet the demand for transportation [Farkas, 2009]. In the Netherlands, for highways and rail infrastructure, the long-term fore- casts regarding the need for new infrastructure is covered in a national vision on in- frastructure and the environment: the Structuurvisie Infrastructuur en Ruimte (SVIR) [Ministry of Infrastructure and the Environment,2012]. In theSVIR, the Amsterdam region is depicted as one of the key-regions, which is of national significance because it comprises for example business district Zuidas, Schiphol Airport, the Amster- dam seaports, and an large expected growth of the housing market [Ministry of Infrastructure and the Environment,2012].
Figure 1.2: Corridor for the OV Schiphol-Amsterdam-Almere-Lelystad (SAAL) project. The project contains improvements and upgrades of the existing rail roads in the area (visualized in green).
1.3 multi criteria decision problems 5
Individual projects of national and international importance are covered in the multi-year plan on infrastructure and the environment: the Meerjarenprogramma Infrastructuur, Ruimte en Transport (MIRT) [Ministry of Infrastructure and Water, 2018]. In the MIRT, Lelystad Airport in the province of Flevoland is depicted to accommodate flights from the national airport Schiphol, since the latter is reaching the limit of yearly allowed flights. Good accessibility of the Lelystad airport and suitable (public) transportation connections between the two airports is of major importance (see for example the OV SAAL project [Ministry of Infrastructure and Water,2018, p. 153 - 155],Figure 1.2. The Dutch government, provinces and related municipalities made agreements on the further improvement of the accessibility of Lelystad airport, via road and public transport. The choice for the location of the Hyperloop test site is therefore also in the province of Flevoland, close to the airport, thus possibly part of a future – commercial – Hyperloop route between the two airports [Arup et al., 2017]. Therefore, there is a large-enough need for the demand of transportation envisioned in the selected region.
However, detailed implementation plans and/or considerations regarding route choice for infrastructures near Lelystad Airport are not included inSVIRandMIRT. Besides, the concept of ”Hyperloop” is not included in those documents since the technology is currently in the testing phase: it is currently not ready to be used for transportation of people and/or goods.
1.3 multi criteria decision problems
The current planning practice in the Netherlands regarding large infrastructure projects for motorways focuses on the area surrounding the new infrastructure, thus next to the new infrastructure itself [Heeres et al.,2012].Heeres et al.[2012] noticed a transition from line-oriented approaches to area-oriented approaches, where the for- mer focuses only on the infrastructure itself and the latter incorporates the environ- ment surrounding the planned infrastructure. Thus within infrastructure planning, also the attention for spatial quality of infrastructure and related spatial projects has gained importance. Spatial quality is the ”outcome of an interaction process which brings stakeholders together, instead of a pre-defined value.” [Macharis et al.,2012].
This makes the location problem for the Hyperloop track in the Netherlands a spatial planning problem in which multiple criteria have to be weighted. This type of complex problems can be classified as multi-criteria decision problems [Malczewski and Rinner,2015]. It are complex decision problems with various important – eco- nomic, environmental, spatial, technical, and social – aspects, there are often var- ious proposed alternatives, and a large amount of stakeholders can be involved.
Stakeholders are ”people who have an interest, financial or otherwise, in the con- sequences of any decisions taken” [Macharis et al., 2012]. How can these type of problems be solved, what approaches are available?
An answer can be found in a sub-field of GIS science, which is dedicated to the integration ofGIS and tools for analyzing and solving those multi-criteria decision problems: GIS-MCDA. GIS-MCDAis a collection of methods and tools for transform- ing and combining the geographic data and preferences to obtain information for decision-making [Malczewski and Rinner,2015].
Recent examples of applying aGIS-MCDAapproach for site selection can be found in literature. For exampleGiuffrida et al.[2019] focused on the possibility of includ- ing Public Participation in quantitative evaluation methods such asMCDA.Terh and Cao[2018] created a GIS-MCDA framework for choosing cycle routes in Singapore, and it incorporates the preferences of various stakeholders into various scenarios.
The authors showed how the comparison of different preferences among three key stakeholder groups – in their case the public, a transport expert, and a government planner for cycling – demonstrates how GIS-MCDA ”enables a logical and compre- hensible way of visualizing how differences in opinion can affect the planning out-
come”, in the context of Singapore [Terh and Cao,2018]. Farkas [2009] discussed another form ofMCDA, utilizing a hierarchical decision tree model to find a location that meets predetermined selection criteria for a metro-rail route. Various objectives were prioritized to find the best location for such an infrastructure [Farkas,2009].
Coutinho-Rodrigues et al.[2011] discussed their implementation of aGIS-based mul- ticriteria Spatial Decision Support System (SDSS) for planning urban infrastructures.
Their goal was to provide decision support in the selection of the ”best” alternative of a set of alternatives, based on multiple evaluation criteria.
Thus aGIS-MCDAapproach assumes a set of alternatives. However, for Hyperloop this set of alternative locations is not identified nor characterized, except for the proposed location for the test-track in Flevoland [Arup et al.,2017]. Thus, the site selection for the route of the Hyperloop is the complex decision problem to be solved.
The advantage of using anGIS-MCDAapproach is that it utilizes the storage and analysis capabilities of GIS and the incorporation of value judgements of decision makers through MCDA [Malczewski and Rinner,2015]. ”MCDA can provide assis- tance in understanding the results ofGIS-based decision making procedures, includ- ing trade-offs among conflicting evaluation criteria/objectives, and then use the results in a systematic and defensible way to develop policy recommendations.”
[Malczewski and Rinner,2015, p.11 and p.328] [Nyerges and Jankowski,2010].
Although the Dutch government has the ambition to invest in transportation in- novations [Ministry of Infrastructure and Water,2019] and a research is conducted on where in the Netherlands a test track for Hyperloop should be located [Arup et al.,2017], it is unknown where a future (test) track of a 40 kilometer Hyperloop between Lelystad Airport and Schiphol Airport should be located. The location problem for such an infrastructure system is characterized as a complex multiple criteria decision problem [Malczewski and Rinner,2015]. Assumed that technolog- ical issues with Hyperloop technology are going to be solved, and assumed that something like the Hyperloop is ”welcome” in the Netherlands: is it possible to find suitable routes for the Hyperloop, utilizing theGIS-MCDAapproach?
1.4 research objective and research questions
1.4.1 Research objective
When the implementation of a new transportation infrastructure in a region is pro- posed the choice for the routing is of high importance. A scientific aspect of the
GIS-MCDAapproach is which criteria to include in the model. Namely, who decides to include a certain criterion in the decision making process, for whom or what is a certain criterion of importance? Choosing which criteria to include in an analysis is also the first step inGIS-MCDA[Malczewski and Rinner,2015]. Further, the method- ology for including the criteria in aGISis another aspect of theGIS-MCDAapproach.
How is the data which represents a criterion quantified in such a way that it can be included in aGIS, while at the same time the data does not lose its quality? And how important is each individual criterion? Another aspect ofGIS-MCDAis how the data is combined and how the data for making a decision is calculated transparently and objectively, i.e. the decision rule [Malczewski and Rinner,2015]. For this study it is finding out which route calculation algorithm is suitable. How does that influence the outcome, i.e. a particular route?
Various people and (local) communities can be affected when a new transporta- tion infrastructure is implemented: positively or negatively. It will affect users of the infrastructure positively, by decreasing transit time significantly. People or communities that can be affected negatively are likely those who have no direct or indirect benefit of the new infrastructure. Before (potential) negative aspects would outweigh positive aspects, which could lead to the cancellation of implementation of a new proposed infrastructure, it is worthwhile to investigate potential scenarios
1.4 research objective and research questions 7
and communicate objectively about it, in the Dutch planning practice. A tool that improves communication between various stakeholders is beneficial for the plan- ning process. This research focuses on the design and implementation of such a tool.
In this research is aGIS-based route planning support framework proposed that in- cludes multiple criteria to answer the question of where in the Netherlands to build Hyperloop infrastructure, taking which factors into account, and whether and how the preferred Hyperloop infrastructure routes change based on the perspectives of different stakeholder groups.
1.4.2 Research questions
The Hyperloop is a new type of infrastructure: for this innovation there is no insti- tutional design. Currently there is no standard or agreed way for decision making.
Therefore,GIS-MCDAis proposed a field of research which could give the solution for the location problem of the Hyperloop in the Netherlands. The research question is:
”Where to place the route of Hyperloop infrastructure in the Netherlands taken into account an origin and destination location, using a GIS-MCDA approach?”
It is a qualitative question, since it includes selecting factors which are important in creating a route for Hyperloop infrastructure. Besides, the research question has a quantitative aspect, because the relevant important factors are quantified and included in aGIS. Relevant sub-questions for this research are:
• What is the Hyperloop? The first sub-question focuses on the proposed Hyper- loop technology and the requirements of such a transportation system, such as technological barriers, the expected user demand, and which cities should be connected? A question that is relevant when the system is used to transport people (and not only goods) is: how does the design of the infrastructure af- fect the user experience regarding shocks and g-forces when speeds are above 1000km/hour? What are the implications for the siting of the route?
• What is the GIS-Multi-Criteria Decision Analysis approach? With this ques- tion are various approaches to selecting optimal infrastructure routes, next to the proposed GIS-MCDA method. Next to that, with this sub-question is the methodology ofGIS-MCDAfor transportation infrastructures covered.
• What are relevant factors in a GIS-Multi-Criteria Decision Analysis approach for choosing optimal routes for a ”Hyperloop” infrastructure to be imple- mented in the Netherlands? This question focuses on the selection procedure for the criteria in the GIS-MCDA approach, taking the Dutch context into ac- count. Which criteria are relevant and which are not, which type of criteria can be used, is data available, does the criterion have a geographical compo- nent, which procedure is used to calculate the weighting factor for the vari- ous criteria, and how are the criteria standardized? [Malczewski and Rinner, 2015]. This selection procedure for criteria is described in a separate chapter (Chapter 3).
• How are the suggested routes evaluated? This sub-question focuses on the evaluation or validation of the results of the methodology. When is a sug- gested Hyperloop route an ”optimal” route? What would be the added value of this methodology when compared to another methodology to identify an optimal route or corridor for the Hyperloop?
1.5 reading guide
Chapter 2discusses the theoretical background of this research. It contains an elab- oration on Hyperloop and a possible institutional design for the governance of Hy- perloop in the Netherlands, methodological frameworks for decision making, the core concepts ofGIS-MCDA, literature reviews on GIS-MCDA, and ten articles on the application ofGIS-MCDAin the field of (transportation) infrastructure route planning.
The chapter ends with an overview of factors that are used in those ten studies, and a conceptual model in which important concepts are connected.
Chapter 3uses the results from the discussion in the previous chapter as input and elaborates on the selection procedure for the criteria, specifically for this case study on Hyperloop routes in the Netherlands. The chapter concludes with a list of 17criteria that are used in the methodology.
Chapter 4elaborates on the methodology for identifying optimal routes or corri- dors for the case study. It includes a description to the study area, specific require- ments for Hyperloop infrastructure, the method for defining importance of criteria, considerations when using data in raster format, and the procedures for calculating and evaluating routes.
InChapter 5are the results of the methodology described and discussed, a pos- sible integration in the Dutch infrastructure planning framework is described, and drawbacks or shortcomings of the methodology are discussed.
Finally, inChapter 6is the research concluded by answering the research question, a reflection on the research, and recommendations for future work.
2 T H E O R E T I C A L F R A M E W O R K
2.1 the hyperloop
As a fifth mode of transportation, the Hyperloop will compete with other transport modes on the intermediate distances, i.e. 300 to 600 kilometers [Leibowicz,2018].
It is estimated that 840 passengers can travel per hour (one direction) on the route Los Angeles - San Fransisco, thus around 7.4 million passengers per year. A trip between Los Angeles and San Fransisco would take around 35 minutes [SpaceX, 2013]. The total costs of the project in California, USA, are estimated at $6 billion and $7.5 billion for the passenger and cargo variants, respectively [Dudnikov,2017; SpaceX,2013].
The availability of Hyperloop as transportation mode will also lead to an increase of competition between airports in a region. Voltes-Dorta and Becker[2018] found that the largest airport in California, Los Angeles International, would benefit most from a Hyperloop service between Los Angeles and San Fransisco as envisioned inSpaceX[2013], because the catchment area of the airport is enlarged due to the availability of the fast Hyperloop service. Namely, passengers from the north of California who initially choose for an airport nearby, could travel to Los Angeles International easier by means of the Hyperloop [Voltes-Dorta and Becker, 2018].
Therefore, the Hyperloop can be seen as a disruptive technology for the current techno-institutional complex of transportation [Leibowicz,2018].
A study on historical data for transportation in the USA shows that in the de- velopment of transportation systems in general the “diffusion of infrastructure pre- cedes adoption of vehicles, which precedes expansion of travel” [Leibowicz,2018].
In other words, availability of infrastructure can be seen as the main ’driver’ for ex- pansion of travel. Early in the technology lifecycle, infrastructure provision should be the major policy consideration [Leibowicz, 2018]. Therefore, ? advises to sup- port a change (or transition) to sustainable mobility effectively by investing public resources in various programs over time in a manner that corresponds with that sequence of diffusion processes.
However,Leibowicz[2018] also mentioned the limitations of the methodology, of which one was that patterns that are observed in the historical data might not be valid for a different spatial, sociopolitical, or temporal context. So it is possible that a Hyperloop transportation system follows a different development path than the systems investigated byLeibowicz[2018] – which were canals, railroads, motorized transportation, and airplanes – and for example does not have to overcome a tech- nological lock-in situation. Especially if Hyperloop is developed first in countries in the developing world, where road and rail networks are limited and environmental problems are more acute, this technological solution can be more urgent and can develop differently [Leibowicz,2018]. AlsoRoss[2016] mentions that it is expected that the first Hyperloop will be located in Asia, Africa, India or the Middle East because in developing countries there is no technological lock-in situation in con- temporary transportation systems [Ross,2016]. The study area for the current study is, however, the Netherlands.
Why would the Hyperloop be a good idea for the Netherlands? Developing the Hyperloop can give a first mover advantage, as mentioned in the advisory report by Arup et al.[2017]. With this advantage, a country which develops the technology can acquire a knowledge position and determine the standards. Besides, it will
9
The Netherlands Developing countries Advantages - Determining the standard
(acquiring knowledge position) - Technology leader - Ensures that the Netherlands
will be connected to an European Hyperloop network
- Acquiring good market position
- Binding R&D frontrunners to the Netherlands, growing up to a development hub
- Income from licenses
Disadvantages
- If the development ends the Netherlands could remain empty handed
- Sunk costs through not using the technology - ”First time” effect:
relatively high investment costs - Imitation by free-riders
Table 2.1: Overview of advantages and disadvantages for the Netherlands and developing countries of being a ”first-mover” with regard to the Hyperloop. Source: Arup et al.[2017].
ensure that the country is connected to an international network of Hyperloop in- frastructure. Disadvantages are the risk that comes with the high investment cost for this transportation innovation: if the development stops a country can remain empty-handed [Arup et al.,2017]. SeeTable 2.1for advantages and disadvantages for the Netherlands of being a first mover with regard to the Hyperloop. The first mover advantages and disadvantages for developing countries are also included in Table 2.1.
At the time of writing (2019) there is not (yet) a decision made if the Netherlands will invest in a Hyperloop track. In this study, however, it is assumed that the Nether- lands will invest in a Hyperloop track, and the route will start at Lelystad Airport and finish at Schiphol Airport. Therefore, this study is a hypothetical case study.
If the Netherlands would never invest in Hyperloop, the proposed methodology of this research could be used for other types of transportation infrastructure in- stead, for example conventional, high-speed, maglev, or lightrail railroad planning, or highway planning.
2.2 institutional design
Implementing a new transportation infrastructure such as Hyperloop in the Nether- lands touches upon the institutional design of the existing planning system and institutions. Institutional design is defined as ”the devising and realization of rules, procedures, and organizational structures that will enable and constrain behavior and ac- tion so as to accord with held values, achieve desired objectives, or execute given tasks”
[Alexander,2005]. Planners are confronted with institutional design if a plan or policy includes new projects or programs, if a plan or policy implementation de- mands the reorganization of existing organizations or new organizations, if existing inter-organizational networks have to be transformed or new linkages have to be created, or if a policy or plan involves amended or new legislation or regulations [Alexander,2005]. Next to a technological design – i.e. that which is visible, see for exampleFigure 2.1– also an institutional design is needed argues [Koppenjan and Groenewegen,2005]. Namely, technological (transport) systems – such as Hyper- loop – also requires an institutional structure that is used to coordinate positions, relations, behavior of parties owning and operating the system [Koppenjan and Groenewegen,2005]. The analysis of the institutional design setting involves the analysis of transactions between relevant actors [Alexander,2005]. So institutional design is about the governance with regard to infrastructure.
Elements of institutional design are: laws, rules & regulations, standards, govern- ments, markets, inter-organizational networks, and organizations. The interactions
2.3 frameworks for decision making 11
that these elements intend to affect are: events, customary behavior, norms, habits, practices, and knowledge or world-views [Alexander,2005]. These public, private, formal and informal arrangements are institutions necessary for a system to func- tion [Koppenjan and Groenewegen,2005]. In other words, the institutional design is the set of ”rules of the game.”
How could the Hyperloop be embedded in an existing institutional design? The answer for now is: such an institutional design needs to be created, together with the development of the Hyperloop technology itself. Namely, for the Hyperloop proposed as a ”fifth mode of transportation” – after planes, cars, trains and boats [SpaceX,2013;Palacin,2016] – as of yet, there is no appropriate institutional frame- work.
Figure 2.1: Future cross section of the variety of transportation infrastructure technologies?
Design for the Hyperloop infrastructure is visualized in pink. Image fromVan Lint[2019].
2.3 frameworks for decision making
The lack of a suitable institutional design for Hyperloop in the Netherlands makes the need for other means to decision-making processes. Various methodological frameworks have been used for the ex-ante and ex-post evaluation of transporta- tion plans and projects. These frameworks can be grouped in two main categories:
single criterion methods (the monetary approach) and multicriteria methods (the non-monetary approach) [Beria et al.,2012]. Cost Benefit Analysis (CBA) belongs to the first, Multicriteria Analysis (MCA) to the latter. In their work,Beria et al.[2012] compared the two approaches CBA and MCA for aiding decision-making with re- spect to projects in sustainable mobility, see tableTable 2.2which summarizes their findings. The methods are not mutually exclusive. Beria et al. [2012] argued that the two approaches could complete each other and joint use can add value to the assessment.
In this research, the MCAapproach is chosen because it is a measure that is not only in monetary terms; Hyperloop technology is not existing (yet) thus costs for the infrastructure are difficult to estimate; and theMCAmethod combined withGIS, using the raster data type, is well-developed for analyzing site-selection problems [Malczewski and Rinner,2015].
The next section desribes the main concepts of MCAfor spatial problems. There- after, four literature reviews on GIS-MCDA methods and methodologies are dis- cussed. That is followed by a discussion of selected studies in which theGIS-MCDA
method for solving location problems has been applied for various infrastructure types. These studies are discussed because their implementation of theGIS-MCDA
MCA CBA
Phases
- Definition of the projects or actions to be judged;
- Definition of judgement criteria;
- Analysis of the impacts of the actions;
- Judgment of the effects of the actions in terms of each of the selected criteria;
- Aggregation of judgments.
- Quantification of relevant effects of an investment;
- Translation of future costs and benefits to present day;
- Scheme with variation of surplus;
- Verify that surplus obtained by some actors exceeds surplus loss paid by others;
- Verify that scheme is marginal.
When - Ex-post; ex ante - Primarily ex ante and possibly ex-post
Where - Micro-scale - Primarily large scale
What - Perception of the effect, including “soft” ones
- Quantifiable and measurable effects (“hard”)
Why - Effectiveness - Efficiency
How many - Multi(ple) criteria and indicators - Single criterion and result Priority/ranking - Input (indications from
decision makers)
- Output (support to decision makers)
Strengths
- Participation and legitimacy;
- Democracy;
- Allows qualitative measures;
- Informal.
- Rigour and rationality;
- Largely formalized;
- Transparency;
- “Common language”, known and used worldwide;
- Easy communication of results;
- Independent from judgements;
- Potentially participative.
Weaknesses
- Potential ambiguity, subjectivity;
- Some components of arbitrariness, especially in the perception of public costs vs. private benefits;
- Risk of double counting;
- Lack of clarity, consistency, accountability.
- Difficult technique, expensive;
- Need of many data, sometimes hardly available;
- Practically impossible to assess
“soft” effects (beauty, personal beliefs, attitudes);
- Equity is not a goal directly assessed, but left to decision maker.
Table 2.2: MCA and CBA compared. Source:Beria et al.[2012].
2.4 core concepts of the gis-mcda approach 13
approach can be useful for the methodology of this research. Besides, the discus- sion of these articles yield a large collection of potential (spatial) criteria to be used in the methodology.
2.4 core concepts of the gis-mcda approach
Malczewski and Rinner[2015] defineGIS-MCDAas ”a collection of methods and tools for transforming and combining geographic data and preferences (...) to obtain information for decision making.” There are a vast amount ofGIS-MCDA methods for selecting a route for infrastructure. The authors distinguished two approaches to solving spatial decision problems: spatial simulation and spatial optimization.
2.4.1 Spatial simulation and spatial optimization
Spatial simulation uses a model of real-world spatial systems to perform various experiments. A range of possible solutions is obtained for the spatial problem, of which then the ”best” solution is chosen.
Spatial optimization models try to find the optimal solution to well-defined spatial decision problems [Malczewski and Rinner,2015]. The difference between the sim- ulation and optimization approaches is the starting point: simulation approaches start with actions and then studies effects on the system objectives by testing vari- ous policies under various external conditions, while optimization approaches start with a definition of system objectives and then specifies actions that will satisfy those objectives at the optimum level. In other words, with simulation is a descrip- tive ”what is” approach used, while with optimization a normative ”what ought to be” approach is used [Malczewski and Rinner,2015].
2.4.2 Value scaling, criteria weighting, and the combination rule
Malczewski and Rinner[2015] distinguishes three main concepts in the procedures for solving spatial multicriteria problems, such as finding possible infrastructure routes for Hyperloop. These concepts are: value scaling, criterion weighting, and a combination rule.
Value scaling is the transformation of raw data to comparable units, including a standardization procedure. It involves defining a standardized metric which can be used to compare various datasets, and which has the same value scale (for exam- ple: 0 = lowest cost, 1 = highest cost). For example, Terh and Cao [2018] used a standardization formula to make the various criteria comparable with each other.
A criterion weight is a value assigned to an evaluation criterion that indicates its importance relative to other criteria under consideration. For example, if the cost of criterion A is much more important than the cost of a criterion B, criterion A will be assigned a higher criterion weight, in such a way that criterion A will have more influence on the outcome. Relatively simple methods for criterion weighting are for example the ranking method, rating method, pairwise comparison, and the entropy-based criterion weight method, while more advanced methods are the proximity-adjusted crite- rion weights, range-based local criterion weights, and entropy-based local criterion weights methods [Malczewski and Rinner,2015]. Regarding criterion weighting, Terh and Cao[2018] conducted questionnaires focused on three key stakeholder groups (the public, transport experts, and government planners) to find the most relevant cri- teria for their routing problem. They used a Likert-type scale, which is a rating method.
Finally, a combination rule is a procedure or method for evaluating and ordering a set of decision alternatives.
2.5 literature reviews on gis-mcda
Four literature reviews that reflect on the field ofGIS-MCDAare included in this the- oretical framework because they discuss which methods and techniques are com- monly used and in which fields of spatial planningGIS-MCDAcan be applied.
Camargo P´erez et al. [2014] conducted a literature review on the application of Multi-Criteria Decision Making (MCDM) techniques for urban passenger trans- portation systems. They found that those techniques have mostly been used to aid decision-making in the long and medium term planning horizons. Also, the Analytic Hierarchy Process (AHP) technique is most often used for the criterion weighting. AHP is a technique that aids decision makers to order priorities. With theAHPmethod a series of pairwise comparisons is utilized in order to reduce de- cision complexity [Durmaz et al.,2019]. Furthermore,Camargo P´erez et al. [2014] observed that economic, logistic, and technical decision criteria were mainly consid- ered in early works. Environmental, social, and land-use criteria have been consid- ered in more recent works. Although Camargo P´erez et al.[2014] their literature review included 86 papers related to urban transportation, only six of those papers focused on location problems for infrastructure.
Zyoud and Fuchs-Hanusch [2017] conducted an analysis to estimate the global research productivity for the topicMCDAand to document growing interest in two
MCDA methods: AHP and TOPSIS. The integration of GIS with AHP for different uses is coined as one of the ”hot topics” in MCDA, and it will stay relevant for applications such as site selection, land use planning, remote sensing, and risk assessment [Zyoud and Fuchs-Hanusch,2017].
The systematic literature review by Mardani et al. [2016] on MCDM techniques customized for solving transportation problems included 89 papers. The focus of their literature review was on application areas such as service quality and trans- portation performance evaluation. Planning of new transportation routes was not an application area. Mardani et al. [2016] found that AHP or Fuzzy-AHP methods were mostly used.
Traditionally, in the field of transport MCDA most applications work towards a common set of criteria, i.e. one value tree for all stakeholders [Macharis et al.,2012].
Macharis et al. [2012] argues that in the context of social decision problems the group of decision makers is not necessarily homogeneous – other than in a business setting where the objective of the group as a whole is the same – and will often have conflicting points of view, therefore a common hierarchy of criteria and weights in
MCDAmethods is not possible. They propose the use of Multi-Actor Multi-Criteria Analysis (MAMCA) for the evaluation of transport projects. The MAMCAmethodol- ogy is an extension of MCDA, with the difference that stakeholders are explicitly taken into account. Moreover, the methodology allows the use of non-numeric and non-monetary values in the evaluation. Macharis et al. [2012] argue that MAMCA
deals better with different points of view of the stakeholders, since the criteria in- cluded in the decision making and the weights of these criteria – which vary from stakeholder to stakeholder – are not aggregated before the MCDA is performed. In- stead, only after performing individualMCDAanalyses, with different sets of criteria and related weights, the results are aggregated. With that aggregation step of the
MAMCAmethodology are various alternatives ranked and strengths and weaknesses of the alternatives are revealed.
Moreover,Macharis et al.[2012] mentions that crucial steps of the methodology are the choice of stakeholders, the criteria set, weights of the criteria, and weights of the stakeholders. Strategic bias regarding those steps must be avoided. Strategic bias in the context of groups decision modeling occurs when individual stakehold- ers provide their preference information to a group decision model which, they perceive, will only improve their own outcomes and not the outcomes of the group [Macharis et al.,2012]. For example,Nadafianshahamabadi et al.[2017] evaluated outcomes of twoMCDA’s for a proposed highway project in Tehran, Iran. OneMCDA
2.6 examples of gis-mcda in infrastructure route planning 15
was conducted with input from technical experts: the other was conducted with input from a sample of community members [Nadafianshahamabadi et al.,2017].
Their findings demonstrated that differences in technical knowledge and values seem to result in different MCDA outcomes. This raises the important question:
whose values and knowledge should be included?
This research, on finding optimal routes for Hyperloop infrastructure between Lelystad and Schiphol, also uses an approach inspired byMAMCA, where a vari- ety of sets of weights for criteria are taken into account. Here, a ”set of weights for criteria” depends on the group to which the stakeholder belongs. Stakehold- ers are consulted via questionnaires and the stakeholder groups are based on the stakeholders’ organization background.
2.6 examples of gis-mcda in infrastructure route planning
Various recent (2001 to 2019) examples of the use of GIS-MCDA in the field of in- frastructure route planning are now discussed. This variety ofGIS-MCDAstudies is inspiration for the design of the methodology for selecting Hyperloop routes. The studies are subdivided in two groups: spatial simulation and optimization. Besides, the discussion yield a list of potential spatial criteria that can be used. After this section, the use of Least Cost Path Analysis (LCPA) for selecting optimum routes is discussed sinceLCPAis a core element in most of these articles.
2.6.1 Applications with the simulation approach
A clean sheet approach
In their methodology for identifying appropriate corridors for a highway route in the eastern part of the USA,Grossardt et al.[2001] combined the ”robust rational choice decision methodology”AHP with the ”rigorous spatial analytic framework”
raster-basedGIS[Grossardt et al.,2001]. In traditionalMCDA ”cost” is the deciding factor: the authors used the spatial equivalent ”impedance”, since it is not neces- sarily a monetary based variable. The impedance variable was composed of the sum in total 50 individual impedance variables. These impedance variables (i.e. cri- teria) where grouped in various ”affinity groups” (i.e. classes). With the pairwise comparison methodology was the relative importance of each criterion per theme identified, resulting in weights for criteria and weights for the affinity groups. Con- sequently, the 50 layers with the data for the criteria were converted to raster grids with a cell-size of 30 meters. For each cell were all values added up, resulting in a cost surface for the whole study area. This cost surface was then used to make a
”least-accumulative-cost distance layer”, representing the total cost from an origin point fo the route to any other point in the study area. Consequently, this layer is used to determine the ”least-cost path” between the origin point and any location anywhere on the raster map [Grossardt et al.,2001]. With LCPA an accumulated cost surface is generated, on which a line can be identified which go from an origin point to a destination point in a study area [Bagli et al.,2011;Durmaz et al.,2019].
There are various approaches to the determination of the least cost path in space, see for exampleStefanakis and Kavouras[1995].
Grossardt et al. [2001] stated that their methodology was “intended as a clean- sheet approach to determining the best corridor when the corridor location op- tions are almost infinite and existing road infrastructure is minimal or nonexistent”.
Thus, the methodology is a type of simulation approach instead of an optimization approach [Malczewski and Rinner, 2015]. Namely, various options are simulated resulting in a selection of possible corridors or routes, after which an optimization
approach can be used to select the most optimal one. For the Hyperloop case in the Netherlands this is also suitable: there are currently no existing routes thus possi- ble corridors have to be identified first, after which the most optimal route can be selected.
A spatial MCDA followed by a non-spatial MCDA
More recently, Karlson et al. [2016] focused on ecological and geological criteria with regard to corridor planning for a railway north of Stockholm, Sweden. Their methodology framework consisted of two parts: a design part and evaluation part.
The design part was a Spatial Multi Criteria Analysis (SMCA) to create three railway corridor suitability (raster) maps, taking into account in total eight spatial factors.
Three different weighting schemes for those factors resulted in three different sce- narios. Consequently, the maps were input for a least cost path finding analysis.
Since the original planning documents for the railway corridor north of Stockholm identified two potential origins and one destination (the airport), the least-cost path analysis was run from those two origins. Therefore, a total of 6 different railway corridors were identified for the study area.
Consequently, in the evaluation part of the framework was the performance of each of the 6 potential corridors assessed by calculating various performance met- rics (e.g. habitat loss, corridor length, etc.) [Karlson et al., 2016]. Although an in-depth evaluation of performance of routes is not the objective for this Hyperloop route-simulation study, it is useful to have a performance metric that should be minimized. For this study, that metric is the sum of the least-cost path.
Spatial constraints, benefits, and costs
Keshkamat et al.[2009] created “a planning system that directly takes into account environmental and socio-economic considerations in selecting alternative routes (. . . )”. With their methodology, the authors generate ”various optimal route alter- natives under different policy visions, in a network of existing roads” [Keshkamat et al., 2009]. Further, Keshkamat et al. [2009] stated that the use of GIS in very preliminary stages of the planning of transport routes has hardly been done. The methodology was applied on the case of the Polish part of the Via Baltica highway project in east Europe. While the work of Grossardt et al. [2001] used a clean- sheet approach, i.e. completely new infrastructure; the approach ofKeshkamat et al.
[2009] utilized only existing roads, thus connecting parts of existing infrastructure into one continuous route for the Via Baltica.
The three main components of their methodology were: criteria and data identi- fication; weighting of criteria and themes; and geospatial data-processing. With the criteria and data identification the relevant criteria were selected and assimilated in a model. Four themes (”affinity groups” in Grossardt et al.[2001]) that cover the various criteria are selected: transport efficiency, ecology, social impact and safety, and economic costs and benefits. Raster maps were used for scores per criterion.
Every pixel in a raster map represents a suitability value for a specific criterion. The options for the value of a criterion are: constraint (absolutely not suitable, thus non- compensatable by a good performance of another criterion or constraint), spatial benefit (the higher the value, the better), or spatial cost (the lower the value, the better) [Keshkamat et al.,2009].
Weighting of the criteria and themes is the second part of the methodology and is based on stakeholder preferences and policy visions. Keshkamat et al. [2009] distinguished different perspectives or political ”policy visions”: equal vision, social vision, ecology vision, and economy vision. By putting different weights on these policy visions or themes, various routing scenarios can be compared. By using the expected value method are weights for those policy vision calculated.
The third part of the methodology, geospatial data-processing, is the combina- tion of the various criteria and weights to generate optimal route maps. Geospatial
2.6 examples of gis-mcda in infrastructure route planning 17
Figure 2.2: Impedence maps for four visions: equal, social, ecology, and economy [Keshka- mat et al.,2009]. These type of raster maps are input for a least-cost path search procedure.
datasets that represent the different criteria (21 in total) and their weights were combined to prepare routing suitability maps for the four distinguished policy vi- sions [Keshkamat et al.,2009], seeFigure 2.2. A ”suitability map” corresponds to the ”impedance map” fromGrossardt et al.[2001]. The resulting raster grids have cell sizes of 1000 meters. This relatively wide cell-size is chosen because it relates to the ”minimum direct impact distance”, i.e. the effect of a road corridor on its surroundings, and the three raster sources used for the case study use a pixel reso- lution of 994 meter to 1000 meter, thus accuracy loss during resampling is avoided.
Consequently, with a line-raster extracting algorithm the line weighted means from each resultant raster (suitability) map was extracted to the road vector layer which represents the existing infrastructure. The Dijkstra shortest path algorithm is used to find the path of least total – ”policy vision”-specific – impedance among that network [Keshkamat et al.,2009].
In the Hyperloop case there is no layer that can represent the existing infrastruc- ture, so this last step is irrelevant for this study. However, an advantage of the methodology by Keshkamat et al. [2009] is that it can be used to assess a prede- termined route, in addition to objectively comparing four policy visions and their optimal routes.
2.6.2 Applications with the optimization approach What if there is no expert knowledge available?
The aim of the study ofKim et al.[2014] was to determine most suitable locations for a new high-speed rail infrastructure in Texas, USA. Their work differs from other research in this discussion because the determination of weights for criteria was not accomplished by consulting experts since the authors noticed that the availability of expert knowledge is limited [Kim et al.,2014]. Instead, the authors used Principal Component Analysis (PCA) to determine which of the criteria should be placed together in (at total of five) groups. PCAwas used to define the groups in a more scientific manner when compared with (subjective) opinions of experts, the authors argued.
Thereafter, the weights for the criteria group are based on choice: which group is most important in a specific scenario? Individual criteria are not weighted in this methodology. Like in the work ofGrossardt et al. [2001],Terh and Cao[2018] and Keshkamat et al.[2009] the weights and criteria together resulted in a cost surface (i.e. “impedance map” or “suitability map”) which is in raster format: each pixel indicates a suitability score based on the established scenarios and relationships between the variables. These cost surfaces were generated for each scenario. Then, a least cost analysis is performed to find the sequence of pixels with least possible scores between two points on the cost surface. The chosen points were locations of the major airports in respectively San Antonio and Austin.
So Kim et al.[2014] did not use expert knowledge or stakeholder input to deter- mine weights for criteria. However, their transparent modeling process, whereby each scenario is calculated, encourages public participation in determining which input variables to include, and deciding which scenario to implement.
Sensitivity analysis on input data
Comparable to the approach of Kim et al.[2014] is the work of Yildirim and Be- diroglu [2019]. The difference is that Yildirim and Bediroglu [2019] consulted in total 35 professionals through interviews and questionnaires in order to determine which criteria and weights to include in theirGIS-MCDAapproach. ”Professionals”
are those who have related work experience on similar high-speed rail engineering projects and who have been involved in inventing solutions for those projects, or those who have an academic background in related fields [Yildirim and Bediroglu,
2.7 least cost path analysis and its use in various applications 19
Source Application Stakeholder groups
Bagli et al.[2011] routeing of power lines
No stakeholders consulted in this research. Instead, neutral, economic, health, and socio-economic ”perspectives” are used, with different weights.
Coutinho-Rodrigues et al.[2011] water supply
system No stakeholders consulted in this research.
Durmaz et al.[2019]
gas pipe planning and evaluation
No distinction made between stakeholders or stakeholder groups.
Farkas[2009]
route/site selection for metro network
An expert group with 5 transportation engineers, 3 mechanical engineers, and 2 economists
Grossardt et al.[2001]
highway corridor alignment
No distinction made between stakeholders or stakeholder groups
Karlson et al.[2016]
design and evaluation of railway corridors
No distinction made between stakeholders or stakeholder groups
Keshkamat et al.[2009]
transport route corridor planning
No distinction made between stakeholders or stakeholder groups.
Instead, four ”policy visions” used: equal vision, social vision, ecology vision, and economy vision.
Kim et al.[2014]
optimizing high-speed rail routes
No distinction made between stakeholders or stakeholder groups
Yildirim and Bediroglu[2019]
high-speed railway route determination
No distinction made between stakeholders or stakeholder groups.
Terh and Cao[2018]
cycling paths planning
Three key stakeholder groups: the public (cyclists n=97 and non-cyclists n=105), transport expert (n=1), and government planner (n=1).
Table 2.3: Overview of stakeholder groups identified in the reviewed articles onGIS-MCDA for (transportation) infrastructure planning.
2019]. Moreover, the work of Yildirim and Bediroglu [2019] included a sensitiv- ity analysis, which indicates how the output of a model changes with variations in input. The applied method was the ”one-at-a-time” technique, which involves modifying input criteria one at a time and consequently observing the effect on the output [Yildirim and Bediroglu,2019;Crosetto and Tarantola,2001].
2.6.3 Involvement of stakeholder groups
In the abovementioned examples of GIS-MCDA for transportation route planning, some authors included a distinction of stakeholder groups. In other words, they took the question ”whose values and knowledge should be included?” [Nadafian- shahamabadi et al.,2017] into account. SeeTable 2.3for an overview of identified stakeholder groups, from the reviewed articles. Only in the work ofFarkas[2009] andTerh and Cao[2018] are explicit stakeholder groups identified.
2.7 least cost path analysis and its use in vari- ous applications
The approaches used for the corridor planning in the above mentioned examples (fromGrossardt et al.[2001],Keshkamat et al.[2009],Kim et al.[2014],Karlson et al.
[2016], and Yildirim and Bediroglu [2019]) all used the “traditional” approach to
LCPA[Stefanakis and Kavouras,1995]. The traditional approach toLCPAcan be used for planning of various linear infrastructures that affect its surroundings, ranging from roads to pipelines [Bagli et al.,2011]. This approach is described inStefanakis and Kavouras[1995] and consists of three main steps:
1. Generation of a friction surface in raster format (map b inFigure 2.3);
2. Generation of an accumulated cost surface, which is a raster map that illustrates the cost of movement to any point X,Y from a point of reference X0, Y0. The